University Links: Home Page | Site Map
Covenant University Repository

Estimating An Optimal Backpropagation Algorithm for Training An ANN with the EGFR Exon 19 Nucleotide Sequence: An Electronic Diagnostic Basis for Non–Small Cell Lung Cancer(NSCLC)

Adetiba, E. and Ekeh, J.C. and Matthews, V. O. and Daramola, S. A. (2011) Estimating An Optimal Backpropagation Algorithm for Training An ANN with the EGFR Exon 19 Nucleotide Sequence: An Electronic Diagnostic Basis for Non–Small Cell Lung Cancer(NSCLC). Journal of Emerging Trends in Engineering and Applied Sciences (JETEAS). pp. 74-78. ISSN 2141-7016

[img] PDF
Download (225kB)

Abstract

One of the most common forms of medical malpractices globally is an error in diagnosis. An improper diagnosis occurs when a doctor fails to identify a disease or report a disease when the patient is actually healthy. A disease that is commonly misdiagnosed is lung cancer. This cancer type is a major health problem internationally because it is responsible for 15% of all cancer diagnosis and 29% of all cancer deaths. The two major sub-types of lung cancer are; small cell lung cancer (about 13%) and non-small cell lung cancer (%SCLC- about 87%). The chance of surviving lung cancer depends on its correct diagnosis and/or the stage at the time it is diagnosed. However, recent studies have identified somatic mutations in the epidermal growth factor receptor (EGFR) gene in a subset of non-small cell lung cancer (%SCLC) tumors. These mutations occur in the tyrosine kinase domain of the gene. The most predominant of the mutations in all %SCLC patients examined is deletion mutation in exon 19 and it accounts for approximately 90% of the EGFR-activating mutations. This makes EGFR genomic sequence a good candidate for implementing an electronic diagnostic system for %SCLC. In this study aimed at estimating an optimum backpropagation training algorithm for a genomic based A%% system for %SCLC diagnosis, the nucleotide sequences of EGFR’s exon 19 of a noncancerous cell were used to train an artificial neural network (A%%). Several A%% back propagation training algorithms were tested in MATLAB R2008a to obtain an optimal algorithm for training the network. Of the nine different algorithms tested, we achieved the best performance (i.e. the least mean square error) with the minimum epoch (training iterations) and training time using the Levenberg-Marquardt algorithm.

Item Type: Article
Subjects: T Technology > TK Electrical engineering. Electronics Nuclear engineering
Divisions: Faculty of Engineering, Science and Mathematics > School of Electronics and Computer Science
Depositing User: Mrs Patricia Nwokealisi
Date Deposited: 18 Oct 2013 15:16
Last Modified: 18 Oct 2013 15:16
URI: http://eprints.covenantuniversity.edu.ng/id/eprint/1668

Actions (login required)

View Item View Item